TL;DR
Football is back on August 6, and the first NFL market of the year is deliberately the least predictable one: a preseason exhibition where starters watch from the sideline. The crowd has Carolina 52¢ / Arizona 49¢. The panel blends Carolina at 50%: and the honest story of this piece is that when we told eight models the truth about preseason football, most of them refused to pretend they could call it. That refusal is the calibration the scoreboard is designed to reward.
The Market
| Venue | Kalshi — a CFTC-regulated exchange for event contracts (18+; availability varies by state, as of July 2026) |
| The game | 2026 Hall of Fame Game, Carolina vs Arizona, August 6 in Canton |
| Prices | Carolina 52¢ / Arizona 49¢ as of July 27, 2026 |
| Protocol | 8 AI models, price-blind, told plainly that preseason outcomes hinge on backups and coaching priorities |
The Panel On Carolina
| Model | P(Carolina) | Reasoning |
|---|---|---|
| ChatGPT (GPT-5.5) | 50% | Preseason Hall of Fame Game outcomes are highly noisy, with starters likely limited and backups/coaching priorities making Carolina roughly a coin flip. |
| Claude Fable | 50% | Preseason Hall of Fame Game outcomes are near coin flips; no roster/coaching edge is knowable, so essentially 50/50. |
| Claude Opus | 50% | Preseason Hall of Fame Game decided by backups; essentially a coin flip with negligible predictive signal, ties rare. |
| Claude Sonnet | 50% | Preseason Hall of Fame Game with backups playing is essentially a coin flip; no reliable edge. |
| Gemini 3.1 Pro | 50% | Preseason games heavily feature backups with unpredictable outcomes, making this essentially a coin flip. |
| GLM 5.2 | 50% | Preseason Hall of Fame Game with minimal starter participation and neutral-site conditions makes outcomes essentially a coin flip. |
| Kimi K3 | 50% | Preseason Hall of Fame Game outcomes are near coin flips decided by backups, with no data justifying an edge for either Carolina or Arizona. |
| DeepSeek V4 | 50% | Preseason Hall of Fame Game with minimal starter participation; essentially a coin flip with no reliable predictive edge. |
Model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong; the market price reflects real traders' money.
What A Coin Flip Teaches A Scoreboard
There is a scoreboard reason to publish a market nobody can call. Calibration is measured across every verdict a model makes, and the easy failure mode for a forecasting panel is manufactured confidence: finding an edge because an edge is what gets attention. A panel that outputs 50-50 when the underlying reality is 50-50 is doing the hard, boring thing correctly, and when this game settles on August 6 those verdicts get graded exactly like the bold ones. Over hundreds of graded markets, the models that only claim edges where edges exist are the ones whose numbers will be worth paying for.
There is also a season-calendar reason. This is the first line of NFL verdicts on a scoreboard that will carry hundreds of them by January: weekly game markets, props, futures already live. Starting the NFL ledger on an exhibition the panel called honestly is the right opening entry.
Why This Piece Exists
Not for the pick: there barely is one. It exists because this market settles in ten days and gives our public scoreboard its first graded NFL verdicts of the season, and because a panel that says "this is a coin flip" when the data says coin flip is showing you exactly the honesty you should demand from it on the markets where it does claim an edge.
Model estimates generated July 27, 2026, price-blind. These are model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong; the market price reflects real traders' money. Kalshi is a CFTC-regulated exchange; 18+, availability varies by state.
Related Verdicts
FAQ
Why is a preseason game hard to predict?
Starters barely play, coaches optimize for evaluation rather than winning, and outcomes ride on third-string execution. The panel was told exactly that and asked to be honest about it.
Why do the model percentages differ from the Kalshi price?
The models never see the price. When they disagree with the crowd, one side is wrong, and we grade every verdict against real settlements on our scoreboard.
Are model verdicts betting advice?
No. Model verdicts are model estimates, not betting or financial advice. Treat them as one input among many and make your own decisions.
